AI: Sasha Luccioni warns about the environmental cost of this new technology. • RFI
Artificial intelligence promises us wonders, but what if its unrestrained use risked making us lose much more than time: our very ability to think? The ease with which we solicit AI has become second nature, from "OK Google" to generative models. We unthinkingly delegate cognitive tasks that once required our full reflection. This practical habit makes us question the real cost of this omnipresence. Every digital interaction has a footprint, visible or not. While the energy consumption of data centers is alarming, another more insidious loss threatens: our intellectual autonomy. By letting AI guide us, would our minds become lazy? The facts, however, are undeniable. Serious studies show it: the more we rely on AI, the more our analytical and problem-solving skills become dulled. Imagine an inactive muscle that atrophies; our brain, that prodigious organ, is no exception. We risk losing that sharpness of mind, that unique ability to forge our own ideas. This is a question rarely dared to be asked aloud. This is not an incitement to burn our computers or to dive back into dusty textbooks, let's be clear. AI has its place, an undeniable value, when used wisely. It is rather about awareness, a call for more thoughtful and intentional use. Let's choose more restrained, more specific, and therefore less energy-intensive AI tools. But above all, let's cultivate a mental hygiene where our brain remains in control. Challenge yourself, question, search for yourselves before entrusting the task to an algorithm. It is by exercising our critical thinking that we keep our cognitive abilities sharp, far from any dependence. True intelligence does not reside in the machine, but in our ability to use it without letting it control us. Let's never forget: our grey matter is our most precious capital; let's not let it dilute in digital ease.
Is AI making us lose our brains?
Artificial intelligence promises us wonders, but what if its unrestrained use risked making us lose much more than time: our very ability to think? The ease with which we solicit AI has become second nature, from "OK Google" to generative models. We unthinkingly delegate cognitive tasks that once required our full reflection. This practical habit makes us question the real cost of this omnipresence. Every digital interaction has a footprint, visible or not. While the energy consumption of data centers is alarming, another more insidious loss threatens: our intellectual autonomy. By letting AI guide us, would our minds become lazy? The facts, however, are undeniable. Serious studies show it: the more we rely on AI, the more our analytical and problem-solving skills become dulled. Imagine an inactive muscle that atrophies; our brain, that prodigious organ, is no exception. We risk losing that sharpness of mind, that unique ability to forge our own ideas. This is a question rarely dared to be asked aloud. This is not an incitement to burn our computers or to dive back into dusty textbooks, let's be clear. AI has its place, an undeniable value, when used wisely. It is rather about awareness, a call for more thoughtful and intentional use. Let's choose more restrained, more specific, and therefore less energy-intensive AI tools. But above all, let's cultivate a mental hygiene where our brain remains in control. Challenge yourself, question, search for yourselves before entrusting the task to an algorithm. It is by exercising our critical thinking that we keep our cognitive abilities sharp, far from any dependence. True intelligence does not reside in the machine, but in our ability to use it without letting it control us. Let's never forget: our grey matter is our most precious capital; let's not let it dilute in digital ease.
AI consumes as much electricity as France.
Your AI consumes as much as France. Prepare yourself for a disturbing truth: artificial intelligence, that powerful digital ally, is also an unexpectedly greedy consumer. Behind the algorithms and fluid interfaces lie data centers, veritable digital factories whose energy appetite is colossal. In 2024, their global electricity consumption is equivalent to that of France as a whole, a staggering figure. Imagine: a single one of these data centers, specifically those dedicated to AI, can gobble up as much energy as a city of 100,000 inhabitants. This astonishing concentration of servers and needs reveals an often-ignored aspect of our digital revolution. The energy footprint of these systems is simply terrible, and electricity is not their only insatiable demand. Cooling the supercharged servers also requires enormous quantities of water, an often-underestimated burden on our precious water resources, as if our machines could "think" without drinking. The demand for AI is exploding, leading to the frantic construction of ever more of these greedy infrastructures. We are rushing to find accessible sources of energy and water, without green options being a priority—a choice that will jeopardize our future. This frenetic race significantly increases our environmental footprint, making the transition to sustainable practices all the more complex and urgent. The lack of transparency from tech giants regarding the actual consumption of their AI models is, in this respect, a flagrant shortcoming and a major obstacle to any progress. Without clear figures, it is impossible to demand true accountability or to envision concrete solutions. It is up to us, as users and citizens, to push for more ethical AI and to favor less energy-intensive tools, because every choice counts. AI is here to stay, but its sustainability will depend on our collective ability to impose an energy diet on it, and fast.
The true cost of your ChatGPT request.
Every time you ask ChatGPT a question, do you really know what's behind your screen? When asked how much a ChatGPT query consumes, I must admit that the numbers are there, but often very limited. Giants like OpenAI or Google share data, yes, but without ever revealing precisely where their models run. They tell you about fractions of joules, a minuscule consumption per interaction. But imagine that multiplied by billions of queries every day, all over the world. That's where the problem takes on a whole new dimension, like a drop of water that, added to others, becomes an ocean of energy. And the trend isn't improving. We are seeing increasingly powerful AI models emerge, capable not only of providing a factual answer but also of reasoning and explaining their processes. Think of generative AI that produces entire videos from a simple text: these feats are energy black holes. We're talking about tens of thousands of attempts before reaching the final result, multiple videos assembled. My observations are clear: there is no transparency on these costs. Major tech groups stubbornly refuse to tell us how much their infrastructure and creations actually cost us in energy. We ask them relentlessly, but without result. Why? Because there are no incentives, no laws that compel them to such clarity. In this frantic race for innovation, where everyone wants to implement AI at full speed, the fear of losing outweighs responsibility. They deploy first, we worry about the consequences later, and that's a calculation that comes at a high cost to us. This situation makes it even more difficult to assess the true environmental footprint of these technologies and to make them more sustainable. It's time to demand accountability, because the future of our planet cannot be a mere adjustment variable in their race for power.
Let's demand transparency from AI giants.
Behind the sparkling promise of artificial intelligence lies a gargantuan hunger, and it is high time to demand that it be held accountable. The world of AI is buzzing, flooding us with innovations that sometimes seem magical. This magic comes at a cost, often invisible: a colossal environmental footprint that tech giants are far too adept at concealing. It is urgent to demand accountability, not from the algorithms themselves, but from those who operate them. These companies hold the true figures of their consumption but often prefer to keep them to themselves or share only incomplete and unrepresentative data. When it comes to energy consumption, companies measure what is local or what suits them, never the complete picture. They present us with tools that "will do everything for you," without mentioning the carbon footprint of their training or their continuous operation. It's a bit like your electricity bill showing only a fraction of your consumption: unacceptable. These data centers are true energy guzzlers, and it is high time to know where this energy comes from – often fossil fuels – and how it is managed, to assess their real environmental footprint. This opacity is unacceptable, especially given the climate challenges. Artificial intelligence operators must give us precise and complete figures, covering their entire chain, from chip production to server management. This would allow us to measure real emissions and push for more sustainable practices. After all, an AI meant to solve complex problems shouldn't create new ones for the planet, should it? Our role as users is not passive; we must adopt a critical stance, as with any new technology. Using AI consciously means asking the right questions: what is the real need, and is there a less energy-intensive alternative? Let's not let enthusiasm blind us, because our individual and collective responsibility is at stake. The promises of AI are beautiful, but its impact demands constant vigilance and thoughtful use. AI promises us the future, but at what cost to our planet if the energy bill remains hidden? It's time to lift the veil and demand unwavering transparency for a truly responsible technology.
The race for the biggest data center is a mistake.
Data centers: Smaller, smarter. The idea that the biggest data center is always the best is driving us straight into a wall, both ecologically and technically. We often talk about data sovereignty, the need to keep our crucial information close to us. For sectors like finance or medicine, it's even a legal requirement to host locally. Yet, the mad rush for giant, centralized data centers seems to contradict this basic logic. We end up with colossal infrastructures, true energy-guzzling cities, whose construction and operation weigh heavily on our environmental footprint. This quest for "bigger" often masks an unflattering reality: astronomical electricity consumption, often powered by fossil fuels. It's a stubborn industry myth to believe that a single behemoth is enough. What if we changed perspective? Instead of dreaming of the single, largest data center in the country, why not consider multiple, smaller, and better-distributed units? Imagine five medium-sized data centers, each integrated locally, taking advantage of existing infrastructures and available renewable energy sources. This decentralized approach offers much more than a simple technical alternative. It allows for better energy management, optimization of green power, and, crucially, it strengthens our digital sovereignty. Gone is the obsession with size for size's sake; intelligence now lies in distribution and efficiency. It's not a question of brute power, but of optimization and common sense. A small cluster of data centers designed for sustainability can outperform a behemoth in terms of impact and resilience. It's about favoring infrastructures that integrate with their environment rather than devouring it. Transparency, often absent among digital giants, would be facilitated by human-scale installations. We could finally know the true energy consumption of our AI tools and demand cleaner practices. After all, our vital applications deserve better than an infrastructure hastily designed for big numbers. The future of AI is local, green, and above all, smart, not just immense.
